Superimposed sleep spindles, resolved by adaptive time-frequency analysis.

This is an example of how adaptive time-frequency parametrization (Matching Pursuit algorithm) can resolve EEG structures overlapping in both time and frequency. 21 channels of sleep EEG were subjected to Matching Pursuit parametrization. Time-frequency atoms corresponding to sleep spindles were chosen to construct this maps of their ime-frequency energy distribution across the derivations.

click here for a larger image

We observe that the faster spindle is more pronounced in parietal, and slower one - in frontal derivations, with transition appearing through Pz, Cz, F7. We may now examine full time-frequency maps and parameters of t-f atoms fitted to EEG by Matching Pursuit for these central electrodes:


EEG data courtesy of I Dept. of Psychiatry, Warsaw Medical School. This page presents results obtained in Laboratory of Medical Physics, Institute of Experimental Physics, Warsaw University. Authors: Dobieslaw Ircha (Java Applet & MP decomposition software), Jaroslaw Zygierewicz (above picture & spindles research), Piotr J. Durka (typing & coffee machine :-)